On Eapid Periodic Variations of Terrestrial Magnetism. 39 



question. For the purpose of a subsequent reference, the vakie 

 of the ratios are also plotted with ^|T, i.e. the reciprocal of the 

 periods as abscissa (the dotted line in Fig. 12). 



The irregular dispersion of the points in Fig. 11 is too 

 remarkable to be considered as due to the inaccuracy of the 

 measurements of the record or to the inconstancy of the 

 instruments, but must be regarded as actually inherent to the 

 nature of the phenomena. Neither is the irregularity^ at all 

 eliminated, even if we take the ratio of the amplitude of Z to 

 that of the resultant horizontal component, ^àX- + AY^ instead, 

 which was actually calculated for the earlier period of the year. 



Though we have not calculated the ratio AZj^ AX^+AY^ 

 throughout the year, it will surel}^ be an overestimation of the 

 Y-component if we take AZj^\AX for it, since, as will be seen 

 later in §15, the azimuth of the horizontal component of the 

 disturbing periodic field is generally less than 45° and about 23*°7 

 on an average (see Fig. 16). 



14. Phase retardation of Z-component with respect to X- 

 component. Next, the phase relation of X- and Z-components 

 were to be investigated. Since the corresponding waves in the 

 two components appear as a rule widely apart on the photographic 

 records where no special zero-line was recorded, a device was 

 necessary to accurately mark off the corresponding time in each 

 of the components at any part of the record. For this purpose, 

 a kind of sliding T-square, originally constructed by Dr. Kadooka, 

 was found very convenient. For the waves of shorter periods, 

 utmost caution was still needed to avoid serious mistakes, very 

 liable to be committed b}^ the slight inclination of the sliding 

 lineal which was to be kept parallel to the time-mark lines. 



The mean retardation of the maxima and minima of AZ 

 relative to those of AIL were calculated in fractions of the periods. 

 The corresponding values for maxima and minima respectively 

 were often sensibl}^ different, especially when the waves are not 

 of simple form, in which case the mean of the two values was 

 simply taken. Plotting tlie values for different trains with the 

 periods as abscissa, the annexed figure was obtained (Fig. 13, A). 



